Abstract:

How do readers integrate incoming words into the local and global discourse context is a central issue in discourse comprehension. Some researchers have suggested that the integration of incoming information with local, proximal context is an automatic and default process. However, the integration of incoming information with the global discourse context employs more attention and may be delayed. On the other hand, some researchers have suggested that incoming information is immediately integrated into the global discourse context. Previous studies also obtained discrepant results concerning the time course of contextual modulation, which might be due to the fact that different types of knowledge violations were used in different studies. Two kinds of knowledge violations were used in previous studies: selectional restriction violations and world knowledge violations. Selectional restriction refers to the semantic constraints on verb’s arguments, which are considered as lexical information. Previous studies have shown that the processing of selectional restriction takes precedence over that of world knowledge. However, it is still unclear whether the precedence of selectional restriction would affect the modulation of global context. Therefore, the current study set out to explore how different kinds of knowledge violations affect global integration in discourse context in an eye-tracking experiment. We constructed discourses consisting of three sentences. The first and second sentences made up a non-fictional or a fictional discourse context. The third sentence was either in (1) selectional restriction violation, (2) world knowledge violation or (3) congruent conditions. In the non-fictional context, sentences in both the selectional restriction and world knowledge violation conditions were incongruent while only the congruent sentences were sensible. However, in the fictional context, all the three kinds of sentences were congruent. 26 university students participated in the experiment. Two were removed due to excessive track losses. Data from 24 (12 males, Mean age = 22.92) participants were used in the statistical analysis. Results showed that for selectional restriction violations, the effect of context was significant in first fixation time, go-past time, total time and number of fixations on the target word as well as first fixation time, total time, number of fixations and regression out count on the post-target region. However, for world knowledge violations, the effect of context was only found in total time and number of fixations on the target region as well as on the post-target region. Participants read the target words and post-target words longer, more frequently in fictional context than in non-fictional context. These results revealed that when object nouns violated selectional restriction, discourse context affected both the early and later stages of word processing. When object nouns violated world knowledge, contextual modulation was only observed on the later time measures. The present study demonstrated that discourse context could override local anomalies, irrespective of the kinds of knowledge that are violated. However, the time course of the integration process was modulated by the kinds of knowledge that are violated. These results establish kinds of knowledge as an important factor for semantic integration in discourse context.

Key words: discourse context, semantic integration, selectional restriction, world knowledge, eye movement.